David Timis: “Sovereignty is a pipe dream. We can hope for less dependence”

Europe needs to focus more on where AI can extend its economic advantage and to become less dependent on the US and China on the layers of the tech stack where it has strengths.

Image: Gemini

David Timis works at the intersection of research and applied AI, with a special focus on how AI shapes the future of work, workers, and society. In his work, he highlights the tensions between what AI might bring to labor markets and the societal impact this disruption has and explores plausible policy responses to these changes.

Work/Code: The conversation around AI policy in Europe seems to have shifted. A couple of years ago, the focus was almost entirely on safety and regulation. Now there’s much more talk about adoption. Is that the right direction?

David Timis: The direction of flight is the right one. For too long, the focus was purely on using regulatory power to constrain American and Chinese tech companies – the main providers across the AI stack, be it cloud, chips, or models. It’s laudable that the Commission has shifted toward “let’s adopt AI”. The question is whether it will be able to move from narrative to action.

There’s actually a historical precedent for this. The printing press was invented in China, but it was scaled and leveraged most effectively in Europe, especially by Johannes Gutenberg. The same happened with the incandescent light bulb which was developed in the UK by British physicist and chemist Joseph Swan, but commercialised in the US by Thomas Edison in the 1880s. Europe has a history of taking technology invented elsewhere and making something of it. But that doesn’t mean we can just be passive adopters. We do need at least the bare minimum in terms of European models and cloud infrastructure, especially for the public sector and critical utilities. On manufacturing leading-edge chips (the EU’s global production capacity in semiconductors is currently below 10 per cent), I think it’s more of a pipe dream to reach the 20 per cent by 2030, which is the figure mentioned in the European Chips Act, but one can at least dream, right?

You make a distinction between sovereignty and dependence. What’s the difference, and why does it matter?

The term “sovereignty” makes some people, especially in the EU bubble, think that we can actually be fully independent, in a sort of autarkic sense. And that’s simply not feasible given the globalised nature of our supply chains. Even with all this rollback of globalisation coming mainly from the US, you can’t decouple decades worth of intricate connections.

The danger is that if we pursue sovereignty, we end up chasing all layers of the AI stack simultaneously, diluting our efforts, when in reality we’d be better off being very focused in the two or three areas where we already have some strengths we could double down on. Europe needs to take a hard look in the mirror to see where it actually stands before deciding what it can do to adapt. If from the first step you decide to pursue something that isn’t really feasible, you’ll go down the wrong path and dilute your already strained resources.

What we should pursue instead is less dependence by building strong alliances with other middle powers such as Taiwan, Japan, South Korea, and the UK, to name a few of the countries that have a stake in the AI race. They’re also dependent on the US and China, but they’re in a better position than Europe because they invested in local digital infrastructure. Sovereignty is a pipe dream. Less dependence is achievable and worth pursuing.

The US launched the Pax Silica initiative; a partnership on chips and AI infrastructure with like-minded countries. Why wasn’t this something Europe initiated?

It was a smart initiative from the US State Department. I just wish Europe had initiated it. That’s exactly what we need: a coalition of like-minded countries coordinating on the chip and compute layer, to name just one layer of the five layer AI stack. The challenge is that it’s being directed by the US. They’re the master puppeteers. And Europe wasn’t even invited as a whole, even though a few individual member states have: the Netherlands (a key partner in the initiative, given that it’s where ASML has its headquarter), Finland, Sweden, and Greece.

That’s actually a perfect illustration of what we should be doing ourselves: initiating similar alliances where the terms are set by a coalition of European member states, Japan, South Korea, and Taiwan for instance. Instead, we’re being invited to join someone else’s table. We’ve missed a beat, yet again. The reason the US can move fast on this is that they’re one country. That’s a huge advantage. The EU cannot move that fast as an imperfect union.

Let’s zoom out to labor. If Europe actually succeeds with AI adoption – what does that mean for the workforce?

The clearest signal of where things are heading is what’s already happening in the US. AI is hitting entry-level work hard, and while the layoffs aren’t all directly AI-caused, some of them are. Companies are running leaner because compute is expensive, and cutting headcount is the quickest way to pay for it, since labor is the largest cost many of these tech firms have. And that’s much easier to do in the US than it is in Europe, given our labor protections.

Then look at Southeast Asia – the Philippines, India, Bangladesh. Their main advantage was cheap labor. Now, a lot of the customer service and business process outsourcing work that Western companies outsourced to the region can be done with AI at a similar price point. They’re being squeezed hard and the plight of workers doesn’t get much media attention.

In AI, there’s real potential for Europe in the healthcare and energy sectors, where European R&D has historically been strong, and in defence.

In Europe, we’re better insulated, at least for now. But I fear it’s just a matter of time until the AI wave hits our labor market as well. And the real challenge for Europe is this: we need to adopt AI to boost economic growth. Without growth, we’re prolonging the inevitable as slow growth can no longer support the standard of living we currently enjoy. And if we adopt AI, we’ll face at least some labor market disruption. We’re stuck between a rock and a hard place. But I do think we’re better prepared than most to deal with the disruption when it comes.

Where should Europe be doubling down? Which sectors offer a genuine competitive advantage?

I see three sectors in particular: First, healthcare. Everyone in Silicon Valley says it, and the VCs in Europe are increasingly saying it too. AI could provide us with an unique opportunity to make healthcare more accessible, to personalize treatment, to unlock cures for diseases that were previously incurable. And Europe, with its strong R&D background in this space, is well-positioned.

Second – and I say this as someone coming from the nonprofit side – defence. It’s not an area we’re traditionally strong in, but there’s real momentum. Helsing for instance is a European defense tech company I like to bring up because globally there are very few companies threading the line between defense and tech the way Palantir or Anduril do in the US. Helsing is doing that in Europe. And as a side note: if Ukraine eventually joins the EU – which I hope happens – they’ll bring an enormous depth of experience in defense tech and AI systems.

Third, green tech. We won’t compete with China on the cost of solar panels or EVs, that’s for sure. But we might be able to come up with our own breakthroughs in the energy efficiency space that AI enables or at least helps facilitate. AI can be not just a drain on our energy sources but also a boon — making grids smarter, energy systems more sustainable. There’s real potential there that aligns with where European R&D has historically been strong.

You’ve pointed out that there’s a disconnect between what economists prescribe and what workers actually want. Workers want job security. Economists recommend reskilling and adjusting to shifts in the labor market. How do you square that?

This really struck a personal chord for me, because I’ve spent ten years working at the intersection of technology, workforce development, and employment. Five out of these ten years I worked for Google, helping equip people, including small business owners, young people and politicians with digital skills, and for the past five years I’ve worked for Generation, one of the world’s largest employment nonprofits, that prepares, places and supports vulnerable people into meaningful careers that would otherwise be out of reach. And so I know first-hand how difficult it is to tell someone who’s vulnerable with limited experience trying to get their foot in the door, that they need to reskill for jobs we don’t even know how they will look like.

A recent study from the Forecasting Research Institute surveyed economists and AI experts, and two things stood out to me. The first one was that even though the economists and AI experts acknowledged a growing chance of transformative AI disruption in the next three to four years, they didn’t expect it to produce significant productivity growth. But we’re already starting to see AI-enabled productivity in the US economy. That disconnect surprised me.

The second was that economists were quite adamant that reskilling would be enough to support people that will inadvertently be displaced during this transition. But most workers say “a certificate is nice, but I need a job guarantee”. If policymakers primarily listen to economists on this and not to workers, we’ll get the diagnosis wrong. Safety nets are key.

What policy tools do we have? Is a Universal Basic Income (UBI) the answer, for example?

Before thinking about UBI, I think there are more targeted tools that are easier to implement and could even prevent the need for UBI. The model I’m most excited about right now are universal learning credits, which for example Singapore has introduced and calls it its SkillsFuture Credit (SFC) system. Since 2015, all Singaporean citizens aged 25 and above receive an opening learning credit of SGD 500 ($370+ USD), which they can use throughout their lifetime to reskill. As of May 2024, Singaporeans aged 40 and above receive an additional SGD 4,000 ($2,900+ USD) to encourage a substantial midcareer skills reboot.

A Universal Basic Capital scheme could help distributing the wealth generated by AI to the wider population, not just financially savvy people.

Still, I wouldn’t take UBI off the table. It could be genuinely useful for categories of workers who often fall through the cracks of traditional unemployment insurance: gig workers, delivery drivers, freelancers, which was made more obvious during COVID. But the biggest challenge with UBI is redistribution. If OpenAI or Anthropic agreed to give 10 per cent of AI gains back – where does this go? To the US government? That would just entrench inequality globally. For redistribution to work at the scale AI demands, it needs to be thought of with a global lense, maybe through a neutral entity, a CERN for AI redistribution. Thinking about the redistribution of the gains of AI on a country or even regional level will just not be enough.

At an individual level, there are three ways through which the average person can benefit from AI right now. You can invest in the companies developing the technology. You can use AI to become more productive at work and hopefully be recognized for it. Or you benefit from AI by way of a shorter work week, but that’s arguably a more utopian option. Of those three benefits, investing is probably the most reliable. A Universal Basic Capital scheme that distributes the wealth generated by AI to the wider population would make that possible for everyone, not just people financially savvy enough to figure out investing works on their own.

Is Europe’s institutional framework an asset in this transformation, or is it too slow?

Our safety nets are definitely better than anywhere else in the world, and they’ll be crucial to support the many, not just the few during this AI transition. But we cannot be complacent about the current situation if we want it to last. Those safety nets are already under pressure – not even because of AI, but because of slow economic growth and demographic trends. We’ve already seen Belgium reduce lifetime unemployment benefits to twelve months in certain cases. That kind of erosion will continue if we don’t improve our economic growth.

I sometimes make a dark joke that the AI race might be won by the country that “breaks” last. But we really don’t want to bet on that strategy – hoping that others like the US or China hit the wall first and we pick up the pieces. We need to compete in order to remain relevant, and preserve the European model, not just rest on our 20th-century industrial greatness.

Our safety nets are already under pressure – not because of AI, but because of slow economic growth and demographic trends.

And yes, institutions help: I’m seeing real improvement in the Commission’s pace over the past couple of years. But often, we’re spreading ourselves too thin: There are several regions in Europe with cheap energy such as France, Spain or the Scandinavian countries (cheaper than the EU average; we’re still the continent with the highest energy prices in the world), but the EU’s AI factories (often referred to as AI gigafactories) are being deployed in a distributed manner across multiple member states to ensure political cohesion, not what makes sense.

What we ultimately need is leadership with genuine vision and a stronger sense of urgency, empowered by member states to make decisions for the EU as a whole, the way Xi or Trump can make decisions for their countries. That might sound like wishful thinking. But it never hurts to dream of a United States of Europe. Because right now, too much of what we need gets blocked by the imperfect nature of the Union, not because Commissioners aren’t trying, but because member states are opposing. And we move, as a result, only like a faster turtle.

tl;dr

Europe does not have to be sovereign, but less dependent: Europe can’t and shouldn’t try to own every layer of the AI stack. A more realistic strategy is building alliances with middle powers (Japan, South Korea, Taiwan, UK).

Europe is caught between a rock and a hard place on labor: Adopting AI is necessary to sustain economic growth – but growth will come with labor market disruption and additional stress to social safety nets.

Listen to what workers want: Economists trust re-skilling to bridge the gap. Workers want job guarantees. If policymakers take their cues from economists alone, they’ll get the diagnosis wrong.

The most promising near-term policy tool is universal learning credits: Singapore’s SkillsFuture model (lifetime learning credits for every citizen) is the kind of targeted, practical instrument Europe could realistically implement.

Europe’s best competitive bet is healthcare, defence, and green tech. These are sectors where European R&D has real depth and where AI can create genuine breakthroughs – not just marginal efficiency gains.